Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Data: $\Omega \subset \mathbb{R}^{n}$ is an open connected (may be unbounded) set, and locally $\partial \Omega$ is aLipschitz graph. $S \subset \partial \Omega$ is measurabel and $H^{n-1}(S)>0.$ The Dirichlet data on $S$ are given by non-negative function $u^0 \in ^{1}_{Loc}(\Omega)$ with $\nabla u^0 \in L^{2}(\Omega)$. The given force function $Q$ is non-negative and measurable.

Consider the convex set \begin{equation} K:=\{ v \in L^{1}_{Loc}(\Omega): \nabla v \in L^{2}(\Omega) \quad \mbox{and} \quad v=u^0 \quad \mbox{on} S\}. \end{equation} We are looking for an absolute minimum of the functional \begin{equation} J(v):= \int_{\Omega}(|\nabla v|^{2} + \chi(\{v>0\})Q^2) \end{equation} in the class $K$.

Definition: We call $u \in K$ a local minimum if for some smal $\varepsilon>0$ we have $J(u)\le J(v)$ for every $v \in K$ with \begin{equation} \|\nabla (u-v)\|_{L^{2}(\Omega)} + \| \chi(\{v>0\}) -\chi(\{u>0\})\|_{L^{1}(\Omega)} \le \varepsilon. \end{equation}

Lemma: If $u$ is a minimum local, then $u$ is subharmonic, hence we can assume that \begin{equation} u(x) = \lim_{r\downarrow 0} \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, \end{equation} where $\oint $ denotes the mean value.

Proof: For non-negative functions $\xi \in C^{\infty}_{0}(\Omega)$ we have

\begin{equation} 0 \le \limsup_{\varepsilon\downarrow 0} \dfrac{1}{2\varepsilon} (J(u- \varepsilon \xi) - J(u)) \le - \int_{\Omega} \nabla \xi \nabla u, \end{equation} that is, $u$ is subharmonic. Then the limit in the assertuion exists for every $x \in \Omega$, and coincides with $u(x)$ for almost all $x$.

Lemma: If $u$ is a minimum local, then $0\le u\le\sup_{\Omega}u^0$.

Proof:For $|\varepsilon|\le 1$ use $u_\varepsilon:=u-\varepsilon \min (u,0)$ and $u_\varepsilon:=u+\varepsilon \min (\sup_{\Omega}u^0-u,0)$ as a first variation.

  1. The lemma suggests that $u$ is subharmonic if for all non-negative $\xi \in C^{\infty}_{0}(\Omega)$ we have \begin{equation} \int_{\Omega} \nabla \xi \nabla u \le 0 \end{equation} I'd like to know what the relation between this definition of subharmoninic and others. For example, by trudinger, if $u$ is subharmonic we have \begin{equation} u(x) \le \lim_{r\downarrow 0} \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega. \end{equation} Please, correct me if I am wrong.
  2. In the last lemma, What is a first variariation?

3.\begin{eqnarray} \limsup_{\varepsilon \downarrow 0}\dfrac{1}{2\varepsilon} (J(u- \varepsilon \xi) - J(u))&=& \limsup_{\varepsilon \downarrow 0}\left \{\dfrac{1}{2\varepsilon}( \int_{\Omega} -2 \varepsilon \nabla \xi \nabla u + \varepsilon^2 \nabla \xi + \chi(\{u-\varepsilon \xi> 0\}) - \chi(\{u>0\}))\right \} \\ &\le & \limsup_{\varepsilon \downarrow 0}\left \{\dfrac{1}{2\varepsilon}( \int_{\Omega} -2 \varepsilon \nabla \xi \nabla u + \varepsilon^2 \nabla \xi \right \}\\ & & + \limsup_{\varepsilon \downarrow 0}\left \{ \chi(\{u-\varepsilon \xi>0\}) - \chi(\{u>0\}))\right \} \\ &\le& \int_{\Omega} -\nabla \xi \nabla u + \limsup_{\varepsilon \downarrow 0}\dfrac{1}{2\varepsilon}\left \{ \chi(\{u-\varepsilon \xi>0\}) - \chi(\{u>0\}))\right \} \end{eqnarray} Am I right here? Why $\limsup_{\varepsilon \downarrow 0}\dfrac{1}{2\varepsilon}\left \{ \chi(\{u-\varepsilon \xi>0\}) - \chi(\{u>0\}))\right \} \le 0$?

If you want the details can be found in the article Alt, H. M. and Caffarelli, L. A. Existence and regularity for a minimum problem with free boundary. J. Reine Angew. Math., 325, (1981), 105–144.. I thank any hint.

share|cite|improve this question
what is $\chi(\{u-\varepsilon\xi\})$? – Mercy King Aug 10 '12 at 23:28
I'm not sure, but I Think is characteristic function of $u-\varepsilon \xi$. – user29999 Aug 11 '12 at 0:52
$u-\varepsilon\xi$ isn't a set! – Mercy King Aug 11 '12 at 1:01
I was wrong, is $\{u -\varepsilon \xi > 0\}$. Haven't you noticed? – user29999 Aug 11 '12 at 13:25
You just changed it! Now it makes sense. – Mercy King Aug 11 '12 at 13:47

1). Note that this is just a weak form of subharmonicity: If $u$ was smooth, then $$ \int\nabla\xi\nabla u = -\int\xi\Delta u, $$ and so the definition coincides with the pointwise definition of subharmonicity (i.e., $\Delta u\geq0$).

Subharmonicity in Gilbarg-Trudinger is $$ u(x) \le \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, $$ and for all sufficiently small $r>0$.

2). Here a first variation means you plug $u_\varepsilon$ into $J(\cdot)$ and use that $J(u_\varepsilon)\geq J(u)$.

3). Your manipulation is correct, except there should be $\varepsilon^2|\nabla\xi|^2$ instead of $\varepsilon^2\nabla\xi$. For the last question, since $\varepsilon>0$ and $\xi\geq0$, we have $u-\varepsilon\xi\leq u$, so $$ \chi(\{u-\varepsilon\xi>0\})\subset\chi(\{u>0\}). $$

share|cite|improve this answer
First, thank you for your answer I accept your anwer 3. However, what I meant in question 1 was by trudinger $u$ is subharmonic when $$ u(x) \le \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, $$ and for all sufficiently small $r>0$. Hence \begin{equation} u(x) \le \lim_{r \downarrow0} \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, \end{equation} if the limit exists, and the article says that if $u$ is subharmonic, then the limits exists and $$ u(x) \le \lim_{r\downarrow 0}\oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, $$ – user29999 Aug 18 '12 at 1:33
I'd like to know exactly too, what is the definition of subharmonic function for them. Seens that is $u \in $(some space) is subharmonic if for all $\xi \in C^{\infty}_{0}(\Omega)$ we have $$ 0 \le -\int_{\Omega} \nabla \xi \nabla u$$ and compare with the definition in Gilbarg-trudinger. I accept your answer 2 too. But I do another ask. How to prove the last lemma. What I have tried was, in $\{u\ge0\}, u_\varepsilon=u$.Hence, $J(u_\varepsilon)\ge J(u)$ implies $$\int_{\{u\le 0\}}|\nabla u_\varepsilon|^2 -|\nabla u|^2 +Q^2(\chi_{\{u_\varepsilon>0\}}-\chi_{\{u>0\}} \ge 0$$ – user29999 Aug 18 '12 at 1:41
and as in $\{u\le0\}$ we have $u_\varepsilon = (1-\varepsilon)u\le0$ because |\varepsilon|\le 1 we have \begin{eqnarray} \int_{\{u\le 0\}}|\nabla u(1-\varepsilon)|^2 -|\nabla u|^2 &\ge& 0 \end{eqnarray} On the other hand we have \begin{eqnarray} \int_{\{u\le 0\}}|\nabla u(1-\varepsilon)|^2 -|\nabla u|^2 &\le& 0. \end{eqnarray} Then $|\nabla u|= 0 $ a.e and $u$ is constant a.e equals to 0 in $\{u<0\}$. This show that $u\ge0$. Is this right? – user29999 Aug 18 '12 at 1:41
@user29999: Yes it is correct. The intuition behind this is that if $u$ gets below zero, you lift the below-zero part slightly so that the gradient gets smaller in that part. – timur Aug 18 '12 at 2:08
I made a typo in my first comment. The correction is: and the article says that if $u$ is subharmonic $$ u(x) = \lim_{r\downarrow0} \oint_{B_r(x)}u \quad \mbox{for} \quad x \in \Omega, $$ – user29999 Aug 18 '12 at 17:37

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.